The AI Is Still Bad?

[Mezmorizor]

Yeah, it's not like the deity AI in civ 4 didn't get 4 free combat units, a free worker, 40% less production needed to build units + 40% less upkeep, 30% more science, 40% less cost for civics and buildings, 20% faster growth, etc. The change to 1-unit-per-tile definitely made the AI struggle at warfare more, and so more bonuses have been needed - no doubt about that. But the AI is looking more competent at core tasks that it struggled with in civ 5/6, and in many ways it is the civ AI with the least major bonuses in the last few games. The combat strength boost is big in civ 7, but the lack of free units is a really big improvement, IMO. In civ 5 and 6, higher difficulty levels are almost always about starting off with a significant disadvantage to the AI and watching the disadvantage disappear as you make up for their advanced start. From what I've seen so far, it's looking like civ 7's AI bonuses aren't hugely focused on a good start (in fact, they might be a little too weak at the start of the game right now), which just allows you so much more freedom in your strategies at the higher difficulty levels. Who knows, there might not even be wonders you just know you'll never get because the AI starts off with 3 cities and you need to do the work to build up that infrastructure before you can start trying to build a wonder

Bottom line is that the AI in Civ 4 is hard to beat and playing a Civ 4 difficulty is equivalent to playing at least 3 difficulties up on unmodded V-VII. Maybe more. It's not rare to find maps literally nobody can beat on Civ 4 deity, and that's still true to this day where people have been at it for 20 years and you would be mocked for talking about core strategies of 15 years ago like "specialist/cottage economy", "city specialization", or "axe rush". Okay, you probably wouldn't be mocked, but specialist/cottage economy being a false dichotomy, city specialization being a bad strategy, and axe rushes being simply bad are commonly known and you would be told to stop reading strategy articles written by emperor players 15 years ago.

The bonuses in 4 are also kind of overly maligned. You're not really getting a true "equal footing" fight until Monarch because the AI is forced to take several suboptimal decisions for the sake of removing oddities and not being horrifically exploitable. The two big, obvious ones are needing to spend the early game building an economy (hence why it starts with defenders) and it turning off research to fully upgrade its army in peacemonger mode so you don't surprise DoW Gandhi and show up to his archer defended capital with riflemen and cannons (also why unit spammers who don't successfully war fall off hard and why they specifically get huge upgrade bonuses). Yes, Deity in particular cheats hard and wouldn't be the monster it is if the bonuses weren't absolutely huge, but nobody asking for hard difficulties is asking for stockfish where you're playing a fair game and the AI is fortnight dancing on your corpse an era ahead of you on turn 60. We're asking for the game to actually have replayability because the AI actually puts up a fight. I personally get little satisfaction from making my number go up higher (or lower), but I get a lot of satisfaction from winning more and more consistently on higher and higher maps. That just hasn't happened in Civ for a long, long time because I could realistically just play them on deity blind and win.

You'll also find that miraculously, 4X games don't actually snowball when the AI knows how to play for win conditions and builds its empire competently. Competent AI is just very rare outside of a select few devs in the 4X genre. The fact that the system where you take other people's stuff also has huge effectiveness upgrades at discrete, large timesteps is in itself a giant anti snowball mechanic. A big part of why I'm so frustrated by the status quo in the civ series AI is that I know I'm only slightly above average at 4X games as Civ IV emperor being my true level shows. I shouldn't be playing Immortal blind and stomping or beating deity within 4 games of the first blind playthrough. I have been from Civ V-VII while also throwing in humankind for good measure. Hell, I don't even care that much if deity stays the same, the AI isn't adjusted at all, and you instead just add 3 more difficulty levels with increasingly absurd bonuses. I just want the game to not be "I'm in a winning position" on turn 30. The only thing off limits to me is if the AI doesn't actually play the game and is instead some omniscient antagonist that attacks you sometimes and you just lose on turn 289 because turn 289 is when deity wins. If I cut them down in mid game, they should be cut down and not a threat. If I start the war by pillaging their strategic resources and killing their production centers, the mop up should be easy. It would still probably be better for them to make the AI somewhat competent though. I'm pretty sure even the more casual players would be put off if they knew how often the AI in V and VI would run out of time before successfully creating a win condition.

After starting a new game, I think I started noticing some patterns with AI issues.
AI really loves wonders. Probably more than everything else. I am couldn't build more than 3 wonders so far, because AI just builds them right away.

This probably it also depends on the "chosen" age focus of the AI.

In my current game, still in antiquity, only Franklin has settled up to a limit. Guess what? He's the only one not busy with spamming wonders.

While the AI is wonder happy for sure, the only competent AI in my first game, Xerxes, was wonder happy AND settled 11 settlements in antiquity. I believe he got 6 wonders. I don't know what's going on, maybe it's hilarious incompetence vs independent people, but something makes the AI just not play the antiquity age at all pretty commonly.

I'd also argue that this wonderspamming is a more pressing AI problem than it simply being bad at the game. If you're going to make wonders a focus for one third of the game (really more because modern age is not much), you should stand a chance to get them at high difficulty levels. Plus it does definitely play a role in the AI being bad because wonders are...whelming in civ VII. Some standouts, but overall whelming. The gates, hanging gardens, and the bugged one whose name escapes me are the only ones I'd actually want to build if I wasn't given glorified victory points for it.

And here I am, slogging through some fairly resilient defenses. Isabella's last stand ultimately was doomed of course (no idea why her best horse is taking a bath in the middle of the fight btw), but this sort of campaign takes much longer than it would in Civ6.

But is that because the AI is playing smart or is it because it's the same tanky 1UPT combat except units are built in 1 to 3 turns? I haven't warred enough yet to say either way yet, but the cheap units are definitely going to be problematic for the multiplayer scene where you'll always have to pillage everything, encircle everything, and slow siege everything because units cost 1 turn in core cities. This is also going to cause the game to run into the issue where only having one build a turn is problematic. I'd be pretty surprised if a builders core cities can't do double a unit on online speed.

More generally, I feel like influence was a mistake. I shouldn't be able to pay influence to tell Augustus to screw off, don't backstab me, and let me be a peaceful builder, but I can. On the positive side, it's also not really interesting because it just becomes other yields efficiently. It wouldn't fix things and people would hate if it's removed at this point, but it is an easy start. I'm sure my game would look different if I had my antiquity age BFF attacking me on turn 20 of exploration. Though I'm also pretty sure he only did that because the game bugged out and thought I was stacking on his borders rather than a true denouncement...

 

[ThunderLizard2]

I don't know what 1 and 0's the AI has to crunch to figure out where to settle, but it's borked. It makes no sense. (and this seems to be something that happens after the patch on release day.) I posted earlier how Rizal tried to settle a spot that was just silly. After that I found a natural wonder right next to his "area". I kept thinking "He'll grab that before I can build a settler and nab it.". Nope...he built a settler and sent it WAY UP NORTH next to Augustus.

This was also broken in Civ VI at launch - why can't they release a game where something as basic as this works properly? AI in general is a big disappointment. Shame on Firaxis!

 

[Noble Zarkon]

More generally, I feel like influence was a mistake. I shouldn't be able to pay influence to tell Augustus to screw off, don't backstab me, and let me be a peaceful builder, but I can. On the positive side, it's also not really interesting because it just becomes other yields efficiently. It wouldn't fix things and people would hate if it's removed at this point

They would, as far as I can see it's generally regarded as one of the best things about Civ 7 giving the player interesting choices and trade offs to make. What don't you like about it?

 

[tedhebert]

For all those posts about how easy deity is, and how the AI isn't competitive... Here's my experience for my first deity game... I was heckled and harassed the whole exploration age and modern is starting the same way, crippling my growth by having to defend constantly.

Here's how the other civs are doing:

I would say they are VERY competitive as you can see. I know not all games go like this one, but nothing has been 'easy' in this game so far. And I'm thoroughly enjoying it. I feel I still have a good chance of getting a win, but it certainly won't be a snooze.

 

[Leefizzy]

I remember seeing someone mentioned the ai mod in 4 that made the ai so hard that people complained about it enough that the diety bonuses got removed.
And when that wasn't enough they dumbed it down.

Also saw that the ai mod guy for 7 mentioned that some things are hard coded, but I forgot where in the forums I saw that in. So we gotta wait on that front.

But I will always believe this when it comes to better ai, a small amount of people actually desire better ai, but most who do, do not realize they don't enjoy only winning 1/# of players of the games they play, and then we'll have the same problem that the civ4 mod had.
Would I enjoy getting my ass beat by ai who never gets tired and are always up for another round? Of course, but I have played with enough people to know that I am in the minority of minorities, most people wanna have fun, have a power fantasy or worse complain that the game is broken/rigged/get sad if they come across a hint of a challenge.

Better AI will only happen if some private company decides to do it as a vanity project, cause they're drowning in money and they got nothing else better to do. Cause we sure as hell won't see from anyone else asides from modders and they're restricted if the game has hard code stuff like 7 does.

 

[BornInCantaloup]

If you're new to a game, you play on Normal.
If you're an expert, you play on Hardcore.

This is what difficulty settings are for.
The trouble starts when you win your first game on Hardcore.

You're meant to lose if you play above your difficulty proficiency. If you don't, something is wrong.

 

[Acken]

The AI being not challenging enough is a combination of poor balance and poor tactics. In essence the AI has 2 main components, its tactical/operation AI (units) and then its decisioning (choices). People often talk about the tactical part because it is the most easy to see when it does stupid stuff but on the decision front the game is so unbalanced that a lot of the choices the AI make are terrible. And it is hard to blame it: the AI doesn't read tooltip. It just chooses based on biases arbitrarily chosen by the AI programmer. And there's only 2 paths to make that better, either the game choices must be better balanced so that the discrepancy between good and bad is narrower or you force the AI to play according to some meta which is imo a bit boring.

Also, part of the challenge problem is that the victories are too simple and don't really depend much on what your opponent is doing. Compare that to civ5 culture CV which if you were too slow you could end up not being able to culturally dominate a civ. Or domination which required you to beat all of them while dealing with multiple fronts because everybody hates you and possibly crippling happiness. Here culture boils down to making explorer asap and domination to pick on some weak civ. Then make a project. Exciting stuff.

 

[Steinernein]

I remember seeing someone mentioned the ai mod in 4 that made the ai so hard that people complained about it enough that the diety bonuses got removed.
And when that wasn't enough they dumbed it down.

Also saw that the ai mod guy for 7 mentioned that some things are hard coded, but I forgot where in the forums I saw that in. So we gotta wait on that front.

But I will always believe this when it comes to better ai, a small amount of people actually desire better ai, but most who do, do not realize they don't enjoy only winning 1/# of players of the games they play, and then we'll have the same problem that the civ4 mod had.
Would I enjoy getting my ass beat by ai who never gets tired and are always up for another round? Of course, but I have played with enough people to know that I am in the minority of minorities, most people wanna have fun, have a power fantasy or worse complain that the game is broken/rigged/get sad if they come across a hint of a challenge.

Better AI will only happen if some private company decides to do it as a vanity project, cause they're drowning in money and they got nothing else better to do. Cause we sure as hell won't see from anyone else asides from modders and they're restricted if the game has hard code stuff like 7 does.

Thread 'RHQ Artificially Intelligent AI Mod'

Feb 11, 2025

1.02 is released, and the newest update

Core AI Behavior Changes:
- Settler units now move more aggressively. Non-moving turn limit reduced from 5 to 1, making settlers much quicker to reach their destinations.
- City founding is now a higher priority operation (Priority 4). The AI will be more focused on expansion.
- AI cities get stronger defenses. Cities will maintain larger defensive forces with specific unit composition requirements.

Military Improvements:
- More balanced army compositions for city attacks:
- Minimum 3 melee units (up from 2)
- Up to 12 ranged units (up from 3)...

• notque
• Replies: 782
• Forum: Released Mods

• 

 

[Mezmorizor]

They would, as far as I can see it's generally regarded as one of the best things about Civ 7 giving the player interesting choices and trade offs to make. What don't you like about it?

1. It is incredibly gamey. This is obviously a game and it feeling so gamey was obviously a conscious decision, but it's just a currency and it feels like a currency. For the vast majority of the game I feel like I'm just trading it for yields because the AI is usually not mad at me.

2. While yes, I am giving things up to do so and it looks like just dealing with wars so you can spend your influence on city states and espionage is better, the fact that I can just bribe whoever to not attack me as long as I'm proactive about it ruins a lot of tension in the game. Not doing this doesn't really help either because the things to do with it that don't make everybody love you are actually stronger, so I'm not really adding tension by doing those instead.

3. Influence is a rare enough yield that it very much so feels like I'm getting my developer mandated every 5 turns diplomatic action rather than doing diplomacy or even really making decisions. I can only really improve influence substantially with influence uniques, and that's a different can of worms.

I will admit that diplomacy is the hardest thing to get right, Civ IV's religion into the world becomes secular and now only cares about what you actually did is the only entry in the series that had a good balance of providing tension and allowing roleplaying, and the rest of IV's diplomacy was bad (especially tech trading, good riddance to that mechanic). V they were just kind of psychopaths that acted erratically. VI was too agenda focused which made the villains especially hamfisted. While the end result for Montezuma in IV and VI is the same, he's going to attack you if he's nearby, VI's "I hate you because you played the game" feels way worse than IV's "I built 20 swordsmen so of course I'm going to use them." Diplomacy also has the unenviable quality that you can't use game theory to guide you because game theory optimal is really not fun to play against a computer in. Cooperate if they want to cooperate. Don't if they don't. Everybody is a computer so everybody will cooperate with everybody unless you choose not to in which case you're everybody's worst enemy.

I will also admit that part of this is that I'm a player who wants to get as close as I can be to winning 100% of the maps I roll rather than getting a higher score or winning faster. 10 times out of 10 I'll choose to play for 30 more turns for a guaranteed wininstead of taking a gamble to win in 5.

 

[Noble Zarkon]

The AI being not challenging enough is a combination of poor balance and poor tactics

I play on the middle levels and don't really care about the AI being a challenger.

I do care about them being competent at core game mechanics - I don't want them crossing the continent to settle their 2nd city 5 tiles from my capital, I do want them to see them consistently at their settlement cap, I do want to see them using Commanders etc etc. At the moment it's all a bit too inconsistent for my taste, like the post above they can be strong but they still sometimes make weird choices.

 

[Pfeffersack]

I remember seeing someone mentioned the ai mod in 4 that made the ai so hard that people complained about it enough that the diety bonuses got removed.
And when that wasn't enough they dumbed it down.

Also saw that the ai mod guy for 7 mentioned that some things are hard coded, but I forgot where in the forums I saw that in. So we gotta wait on that front.

But I will always believe this when it comes to better ai, a small amount of people actually desire better ai, but most who do, do not realize they don't enjoy only winning 1/# of players of the games they play, and then we'll have the same problem that the civ4 mod had.
Would I enjoy getting my ass beat by ai who never gets tired and are always up for another round? Of course, but I have played with enough people to know that I am in the minority of minorities, most people wanna have fun, have a power fantasy or worse complain that the game is broken/rigged/get sad if they come across a hint of a challenge.

Better AI will only happen if some private company decides to do it as a vanity project, cause they're drowning in money and they got nothing else better to do. Cause we sure as hell won't see from anyone else asides from modders and they're restricted if the game has hard code stuff like 7 does.

While I don't know if that story about the modded Civ4 AI being dumbed down in the end is true, without a doubt both the foundation Civ4 delivered and the modding result were quite good. And it is also telling that the 4X game with the -IMO- best AI (Old World) is in many aspects fairly close to Civ4 (and its spin-off colonization) - with the big difference of having moved to 1 UpT as well. If you haven't done, I would recommend you to give it a try, as it delivers exactly whyt you describe...a restless AI exploiting the players errors, while it doesn't make too many glaring ones itself. It is even without any boni a fearsome opponent you have to overcome and it took me several hundred hours of playing OW to reach a point where *I* feel competitive to it in that sense.

 

[uppi]

without a doubt both the foundation Civ4 delivered and the modding result were quite good.

The base game civ 4 AI was as terrible as for any civ game. It was only when an AI mod was (partially) integrated into civ 4 (IIRC in BtS) that the unmodded AI became any good.

 

[Melkus]

The base game civ 4 AI was as terrible as for any civ game. It was only when an AI mod was (partially) integrated into civ 4 (IIRC in BtS) that the unmodded AI became any good.

That's how I remember it, too.. But Soren Johnson put more emphasis on AI-friendliness in game design and mechanics than later entries in the series did (though CivVII is much better than the last entry in this regard).

And FXS' commitment to open modding tools was much stronger at the time. Civ4 AIs also had more personality and even unique quirks.

 

[OmegaDestroyer]

As with the other titles, I find the AI lackluster in Civ 7. The only thing that makes it competitive are the cheats. I have fought many battles where the AI could have easily won had it pressed its advantage, but it ALWAYS withdraws when its units start to suffer damage. I fended off Xerxes 10+ unit army, on immortal, with two swordsmen and a slinger. I nearly forgot I also had a city state raiding me. The AI could have easily eliminated me from the game, but retreated after a few units took damage. I was able to build up my army, counterattack, and eliminate him.

Just once, I'd like to see Firaxis make a serious attempt to make competitive AI.

 

[ThERat]

The base game civ 4 AI was as terrible as for any civ game. It was only when an AI mod was (partially) integrated into civ 4 (IIRC in BtS) that the unmodded AI became any good.

That's just not true. Even the base game was no walk in the park on deity. I played it extensively and we played always war, even on emperor the AI was not easy to beat because it did not simply send its units to your land to die mindlessly.

 

[Melkus]

That's just not true.

I unearthed a thread from early Civ4 (still after the first expansion, so not a totally fair comparison) about that if anyone's interested.

Thread 'GOTM Difficulty Levels'

Aug 23, 2006

I am curious what GOTM players think are the reasonable difficulty levels and what they would prefer for GOTM competition.

Should GOTM go up to Immortal or even Deity level. After playing my first complete game at Emperor for GOTM9, I am definitely ready for a break and would be happy to go back down to Warlord. I think cycling between Warlord and Emperor works well. The lower levels give a playable game for new people and the opportunity for the experienced to win big. Of course I would play Immortal next month if that was what is up, but I do not think it would be fun. I got tired of...

• ngraner42
• Replies: 83
• Forum: Civ4 - Game of the Month

The jump from Monarch to Emperor - two AI settlers - was really tough, IIRC from my personal experience. But comparing with 1 UPT systems is never really fair - those stacks were so dangerous, while you can always pick off the unit carpets one by one.

 

[stealth_nsk]

I unearthed a thread from early Civ4 (still after the first expansion, so not a totally fair comparison) about that if anyone's interested.

Thread 'GOTM Difficulty Levels'

Aug 23, 2006

I am curious what GOTM players think are the reasonable difficulty levels and what they would prefer for GOTM competition.

Should GOTM go up to Immortal or even Deity level. After playing my first complete game at Emperor for GOTM9, I am definitely ready for a break and would be happy to go back down to Warlord. I think cycling between Warlord and Emperor works well. The lower levels give a playable game for new people and the opportunity for the experienced to win big. Of course I would play Immortal next month if that was what is up, but I do not think it would be fun. I got tired of...

• ngraner42
• Replies: 83
• Forum: Civ4 - Game of the Month

The jump from Monarch to Emperor - two AI settlers - was really tough, IIRC from my personal experience. But comparing with 1 UPT systems is never really fair - those stacks were so dangerous, while you can always pick off the unit carpets one by one.

Yeah, in Civ1-Civ4 AI was popping out units out of nowhere, but since it happened outside of player visibility, it didn't feel like cheating. And AI was able to just direct those unit stacks toward the target without any complex logic.

 

[Melkus]

Also Civ4's difficulty puts crunching penalties on the player. For comparison, it would be like playing with a lowered settlement limit (say, one per level). It was possible to completely fail at empire building and fall apart at higher levels (even without military pressure).

 

[ThERat]

But comparing with 1 UPT systems is never really fair - those stacks were so dangerous, while you can always pick off the unit carpets one by one.

That's very true and that's why starting with Civ5, the AI's deficiencies became terribly obvious. I agree that Civ 5 and Civ 6 AI was terrible at start. I recently played a Civ 6 game with all expansions and can't say, I was impressed with the AI. I never played Civ 5 VP, so I can't comment on it.
Wasn't Civ 7 branded at easier for the AI due to less snowballing? I doubted it from the start cos I can tell that the AI can't deal with all the convoluted choices and paths.

 

[Melkus]

Vox Populi is / was the best tactical AI for unit carpets I've seen. (Haven't played for a while, it changes all the time and regresses sometimes.)
I actually often fell behind in never-ending wars. They were relentless even on a level like King (but not unfair).

But it relied on (better timed) AI bonuses too - of course.